Research On Numerical Simulation Method Of Aerodynamic Characteristics Of Rotor Airfoil Based On Neural Network

With the rapid development of computer technology and numerical methods,rotor aerodynamic analysis methods have developed rapidly.Although the accuracy of numerical analysis technology is getting higher and higher,the calculation efficiency has not improved significantly.The existing CFD methods are limited by the contradiction between its accuracy and efficiency,and it is difficult to achieve rapid or even real-time prediction of rotor aerodynamics and flow field,so other methods must be used.At present,the new generation of information technology revolution with artificial intelligence as the core is affecting all aspects of society.Experts and scholars have begun to use this technology to conduct exploration and research in aerodynamic modeling analysis,aircraft optimization design,and wind tunnel intelligent control.Given the powerful ability of artificial intelligence to calculate and analyze regression,this article uses artificial intelligence methods to study the aerodynamic coefficient prediction,flow field prediction,and grid generation methods of helicopter rotor airfoils to reduce CFD calculation time.In this paper,three sets of neural network models will be established to predict the aerodynamic coefficients of the rotor airfoil,the aerodynamic characteristics of the flow field of the airfoil,and the grid generation method.Firstly,the domestic and foreign CFD calculation methods and the development status of artificial intelligence technology are introduced,and the research ideas and main research work plans of this paper are explained.Then a set of airfoil mesh generation methods are established,and based on the generated mesh,a high-precision CFD calculation method suitable for airfoil flow field calculation is established,and the validity of the established numerical simulation method is verified.Then introduce the artificial intelligence method and artificial neural network model,and use a classic convolutional neural network to verify the classification and clustering of the convolutional neural network for the non-linear relationship between data and powerful computing power.The deep neural network(CNN + BP)is combined with the CFD calculation method to establish a sample expert database and build and train a neural network model.The influence of the model on the predictive performance of the aerodynamic coefficient is studied from the shape parameters such as airfoil thickness and camber,and the results are analyzed.Then use the CFD calculation method to establish a flow field database,build and train a convolutional neural network(CNN)for airfoil flow field prediction,predict the flow field of different types of airfoils at different Mach numbers and angles of attack,and give prediction Result analysis.Then,a convolutional neural network,RBF neural network(CNN + RBF),and rotor airfoil mesh generation method were combined to develop a neural network algorithm for automatic mesh generation,which reduced the mesh generation work and calculation load.Finally,the methods used in the research process,the advantages and disadvantages of the research results are summarized,and further development prospects are discussed.